9 Steps to Set Up GA4 Lead Tracking and Keep Every Source
Nine steps to set up GA4 lead tracking so every lead reaches your CRM with its source, with a filterable step matrix, a missing-source calculator and CAD prices.

Published 6 March 2025 · Last updated and fact-checked November 2026
Short answer: GA4 lead tracking works when every form records where the person came from and hands that answer to your CRM with the lead. Run these 9 steps in order. Start with one lead event, generate_lead, and end with a weekly report, because last-click credit can differ sharply from models that follow the whole path (Li and Kannan, 2014).
Most installs I open on client sites stop halfway. The Google tag is on the page. Form submissions get counted. The source never reaches the contact record, so nobody can say which channel sent the lead.
Picture a head of marketing at a mid-sized Vancouver consulting firm (a composite, not a named client). In the Monday meeting she is asked which channel produced last quarter’s pipeline. GA4 shows her a count. The CRM shows her names, each with a lead source someone typed in by hand, weeks late. Neither screen answers the question.
This is not a reporting problem. It is a wiring problem, and it gets fixed in a set order. The order here is the one I use for GA4 lead tracking on WordPress and HubSpot sites. Every GA4 menu path matches Google’s help pages as checked in November 2026.
“It is a wiring problem, and it gets fixed in a set order,” wrote Joshua Jackai, Vancouver web designer.
Key takeaways
- Google’s recommended lead event is generate_lead, marked as a key event, which is what GA4 now calls a conversion (Google Analytics Help).
- UTM values are case sensitive, so LinkedIn and linkedin become two sources in GA4 (Google Analytics Help).
- On the Y Scouts Hiring on Purpose funnel, 166 of 938 visitors became leads (17.7%), read from the client’s own HubSpot.
- Consent experiments with more than 80,000 users found small banner decisions changed what people chose (Utz et al., 2019).
- Lead Source Reporting from Jackai Agency, my Vancouver practice, is CA$2,600 and two weeks (price table, November 2026).
| # | Step | Where it happens | Done when |
|---|---|---|---|
| 1 | Decide what to record | On paper | 4 fields are named |
| 2 | Property and data stream | GA4 Admin | One tag, filters in Testing |
| 3 | One lead event | GA4 Admin or the form | generate_lead is a key event |
| 4 | UTM naming convention | A shared document | Every value is lowercase |
| 5 | Source into the CRM | Forms and CRM | Hidden fields fill on a test lead |
| 6 | AI referrals | GA4 channel groups | AI assistants sits above Referral |
| 7 | Test end to end | Phone and CRM | The test source reaches the contact |
| 8 | Watch for drift | Monthly check | GA4 and CRM gap is steady |
| 9 | Weekly report | One page | Leads, qualified, closed by channel |
In this guide
Why does GA4 lead tracking lose the source of a lead?
GA4 lead tracking loses the source because GA4 and your CRM record different things and nobody joins them. GA4 knows where a session came from. The CRM knows the person. Unless the form carries the source across at the moment it is sent, the lead arrives with no channel attached, and someone guesses it later.
On my lead attribution method page I list where B2B tracking usually fails: links go out without UTM parameters, the form drops the tags before they reach the CRM, nothing records the converting page, and AI referrals get filed as direct. Fixing all four is about two weeks of work and no design change.
Google also sets a hard line on what Analytics may hold. Its policy bars sending personal information such as email addresses into GA4 (Google Analytics Help), so the named record has to live in the CRM. GA4 shows which channels work. The CRM shows the lead source on every contact. You need both, joined by the same values.

1. What does GA4 lead tracking need to record?
GA4 lead tracking needs four things recorded for every lead: that a lead happened, where the person first came from, which page they converted on, and a link that attaches all three to the CRM contact. Miss one and the chain breaks there. Buyers reach firms through many touchpoints across several channels (Lemon and Verhoef, 2016).

- The event: one lead event, generate_lead
- First touch: original UTM values, referrer and landing page
- Converting page: the URL the form sat on
- The join: hidden form fields mapped to CRM properties
- Time: about 20 minutes to write down
First touch matters because buyers rarely convert on the visit that found you. Someone finds you through a LinkedIn post, returns through Google a week later, and converts from a bookmark. Keep only the last source and the post gets no credit, so next quarter’s budget moves away from the channel that started the conversation.
“Customers now interact with firms through myriad touch points in multiple channels and media, and customer experiences are more social in nature,” wrote Katherine Lemon and Peter Verhoef (Journal of Marketing, 2016).
That is why I write the four fields down before anyone opens GA4. The list becomes the spec for every later step: the event in step 3, the UTM values in step 4 and the hidden fields in step 5. GA4 lead tracking that stops at a count cannot name a person, and it should not. The count tells you volume. The four fields tell you which channel earned the lead and which page did the persuading, and that is the answer the Monday meeting is asking for. Write the list where the whole team can find it.
2. How do you set up the GA4 property and data stream?
Set up one GA4 property with one web data stream, the automatic events left on, the Google tag installed once, and internal traffic filtered in Testing state first. Check an existing property against the same list. Google lets you raise event data retention from 2 months to 14 months (Google Analytics Help).

- Property: Admin, Create, Property, with the lead generation objective
- Stream: Admin, Data collection and modification, Data streams, Add stream, Web
- Tag: once, through the theme, Google Tag Manager or the CMS, never two of them
- Retention: event data at 14 months
- Time: about an hour
Keep the measurement switch under Events turned on. It records page views, scrolls, outbound clicks, file downloads and form interactions without code. Install the Google tag once, right after the opening head tag on every page, or add it through Google Tag Manager or your CMS. Never both, or every count doubles.
Then define internal traffic. On the stream page, open Configure tag settings, click Show more, then Define internal traffic, and add your office IP addresses. Open Data filters, click Create Filter and choose Internal Traffic (Google Analytics Help). In the same tag settings, click List unwanted referrals and add any booking or payment tool that sends people back to your site (Google Analytics Help), or it shows up as the source of your leads.
Run the internal traffic filter in Testing state for a few days before you switch it to Active. An active data filter is permanent. Anything it excludes is never processed, and you cannot get it back.
“An active data filter is permanent,” wrote Joshua Jackai, Vancouver web designer.
3. Which event should count as a lead in GA4?
Count one event as a lead: generate_lead, Google’s recommended event for a submitted form, marked as a key event. Do not count form_submit, which fires on every form, including the newsletter box. One lead event keeps GA4 and your CRM counting the same people; a lead_source dimension takes 24 to 48 hours to appear (Google Analytics Help).

- Event: generate_lead (Google for Developers)
- Optional parameters: lead_source, value and currency
- Counting method: once per session for a lead form
- Never count: form_submit on its own
- Time: 30 minutes, plus 24 to 48 hours for dimensions
On sites I audit, counting all form submissions as leads is the most common reason GA4 shows more leads than the CRM got. That is my experience, not a published benchmark. Pick how to fire generate_lead by how your form confirms a submission.
| How the form confirms | How to fire generate_lead | What breaks it |
|---|---|---|
| Sends the visitor to a thank-you page | Create the event in GA4 without code, from page_view on that URL | Bookmarked or indexed thank-you pages |
| Shows a message on the same page | Push the event from the form plugin’s success hook, or a Tag Manager trigger | A plugin update that renames the success message |
| Embedded from another platform, such as HubSpot or a booking widget | Use that platform’s own submit callback or GA4 connection | A form in a frame on another domain, which your tag cannot see |
For the no-code route: in Admin, under Data display, click Events, then Create event. Name it generate_lead, switch on Mark as key event, choose Create without code, pick page_view, enter the thank-you page URL, and click Create (Google Analytics Help). Then set that page to noindex, so search visitors cannot trigger it.
GA4 asks for a counting method: once per event, which Google recommends, or once per session. For a lead form I pick once per session, because a nervous double click on Submit should not become two leads. If you send lead_source, register it under Data display, Custom definitions, as an event-scoped custom dimension.
4. How should you name your UTM parameters?
Name UTM parameters with one written, all-lowercase convention that everyone who ships a link follows. Always set utm_source, utm_medium and utm_campaign, and use the medium values GA4’s default channel rules expect. Values are case sensitive, so LinkedIn and linkedin become two sources (Google Analytics Help).

- Always set: utm_source, utm_medium, utm_campaign
- Case: lowercase only
- Mediums GA4 expects: email, social, referral, and anything starting with paid for ads
- Never tag: links between pages of your own site
- Time: an hour to agree, minutes per link after
A tagged link looks like this: yoursite.com/guide?utm_source=linkedin&utm_medium=social&utm_campaign=hiring-guide. Here is the shape I start clients on.
| Where the link goes out | utm_source | utm_medium |
|---|---|---|
| Unpaid LinkedIn posts | social | |
| Paid LinkedIn ads | paid_social | |
| Monthly newsletter | newsletter | |
| Partner site or podcast show notes | the partner’s domain | referral |
| QR code on event material | the event name | qr |
Look at the QR row. GA4 has no default channel for a medium called qr, so those visits land in Unassigned unless you add a QR channel to a custom channel group (Google Analytics Help).
Never put UTM parameters on links between pages of your own site. On my projects that rule has no exceptions. An internal tag tells GA4 the visit came from your own campaign, which hands the credit for the lead to you instead of the channel that brought the person in.
Put the convention in one shared document and link it from wherever links get made: the newsletter tool, the ad accounts, the social scheduler. When a new person joins, that page is the first thing they read about links. I review the GA4 source list once a month for stray capitals or new spellings, and fix the link at its origin rather than patching the report. A convention that lives only in one person’s head breaks the week that person goes on holiday.
“Never put UTM parameters on links between pages of your own site,” wrote Joshua Jackai, Vancouver web designer.
5. How do form submissions carry the lead source into your CRM?
Form submissions carry the lead source into the CRM through three parts: a first-touch script that stores the source on arrival, hidden form fields that send it on submit, and a CRM property for each field. On the Y Scouts funnel, 166 of 938 visitors became leads, each arriving as a contact with a source.

- Capture on arrival: UTM values, referrer, landing page and any ad click ID, stored in the session
- Pass on submit: hidden fields on every form
- Map in the CRM: one contact property per hidden field
- Record: the converting page
- Time: half a day per form type
The first-touch script means the source survives a visitor who reads three pages before filling anything in. In HubSpot, each hidden field maps to a contact property you create for it. Keep both touches, so the lead source travels with the person from first visit to form. GA4 shows the same split in two reports: First user source in User acquisition, and Session source in Traffic acquisition. Report first and last touch side by side instead of letting one model choose for you.
Y Scouts, an executive search firm in Scottsdale, Arizona, shows why. Its lead pages had no campaign tracking, so the founder could not tell which channel produced a lead. I rebuilt the Hiring on Purpose book funnel so the form, the delivery email, the follow-up workflow and UTM tracking run as one HubSpot flow. A submission now becomes a contact with a source.
That funnel converts visitors to leads at 17.7%, 166 of 938, read from the client’s own HubSpot dashboard rather than from GA4. In Q3 2025, with nothing new filmed, the wider content work contributed to inbound MQLs rising from 17 to 30. The tracking did not cause that rise. It is how the founder could see where the new leads came from. The full build is in the Y Scouts case study.
“The tracking did not cause that rise,” wrote Joshua Jackai, Vancouver web designer.

6. How do you see AI referrals in GA4?
See AI referrals in GA4 with a custom channel group that lists the assistants by source and sits above Referral. GA4’s default group now has an AI Assistant channel for sources like ChatGPT, Gemini and Copilot, but Perplexity is not on Google’s list (Google Analytics Help). Standard properties allow two custom groups.

- Channel name: AI assistants
- Condition: Source matches regex chatgpt|openai|perplexity|claude|gemini|copilot|deepseek|grok
- Order: above Referral
- Applies: retroactively, per Google
- Time: 15 minutes
Google’s documentation says the AI Assistant channel leaves out Google’s own AI Overviews and AI Mode. An assistant that sends no referrer shows up as Direct; my lead attribution setup infers those from landing page and session pattern, labelled probable, not certain.
To build the group: in Admin, under Data display, click Channel groups, then Create new channel group. GA4 starts you from a copy of the default group. Click Add new channel, name it AI assistants, add the Source condition, then click Reorder and drag AI assistants above Referral, because GA4 files each visit in the first channel it matches (Google Analytics Help). Click Save group. Add the same AI channel to your CRM’s lead source picklist, so an AI-referred lead carries one label in both places.
Why bother? Assistants are now a place where buyers meet your firm, and the GEO study found content changes could raise visibility in generative engine answers by up to 40% (Aggarwal et al., 2024). The same paper warns that site owners have little control, so measure what arrives instead of assuming it.
“Content creators have little to no control over when and how their content is displayed,” wrote Pranjal Aggarwal and colleagues (KDD, 2024).
So give those leads their own line. My free AI Visibility Report checks whether assistants mention your firm at all, in 24 hours.
7. How do you test GA4 lead tracking before you trust it?
Test GA4 lead tracking the way a real lead would arrive: open a tagged link on your phone’s mobile data, browse two pages, submit the form, then open the new CRM contact. Every source field should hold a value. Google’s DebugView shows the events live while you do it (Google Analytics Help).

- Test link: ?utm_source=test&utm_medium=email&utm_campaign=tracking-check
- Device: a private window on mobile data, outside the internal traffic filter
- Debug: Google Tag Assistant plus Admin, Data display, DebugView
- Pass: generate_lead in Realtime and every CRM source field filled
- Time: about 10 minutes per form
Here is the order I run before I call a setup done. Open the tagged link. Connect the site in Google Tag Assistant, then open DebugView, and accept analytics cookies, or DebugView stays empty. Visit two other pages, then submit the form; the source should still read test. The Event count by Event name card in Realtime should show generate_lead within a minute or two.
Then open the CRM record. Every lead source field on the new contact should hold a value: test, email, tracking-check, plus the landing and converting pages. Last, check the thank-you URL. If the address contains the person’s email or name, fix the form before launch.
Google requires page URLs and their parameters to be free of personal information (Google Analytics Help), and a thank-you page with ?email= in the address sends that address with every page view. GA4 can redact email addresses on a best-effort basis. The reliable fix is a form that never puts them there.
8. Where does GA4 lead tracking break after launch?
GA4 lead tracking breaks after launch in five predictable places: consent banners, duplicate tags, untagged links you own, rebuilt forms that drop hidden fields, and forms people abandon. A setup that passes on day one can drift over time. In experiments with more than 80,000 users, banner design changed consent choices (Utz et al., 2019).

- Consent: visitors who decline analytics cookies are partly or fully missing
- Duplicate tags: a plugin and a theme both adding the Google tag
- Untagged links: email signatures, PDF guides, podcast bios
- Rebuilt forms: a new plugin that drops the hidden fields
- Check: monthly, about 20 minutes
How many visitors accept depends on the banner, not only on the visitor.
“Seemingly small implementation decisions can substantially impact whether and how people interact with consent notices,” wrote Christine Utz and colleagues (ACM CCS, 2019).
Canadian firms have a legal side too. Analytics cookies and form data fall under the federal Personal Information Protection and Electronic Documents Act (PIPEDA), or a provincial equivalent such as BC’s, so the banner wording is not only a tracking decision. My post on website lead tracking in Canada covers the consent setup.
The form itself matters as well. Tracking a form people abandon only measures the abandonment more precisely. In an eye-tracking study, forms rebuilt to usability guidelines were finished faster with fewer failed submits (Seckler et al., 2014).
GA4 and the CRM will never match exactly. Treat the CRM as the count of people and GA4 as the map of where traffic came from. On the sites I have worked on, a small, steady gap is normal. Write down the gap in the week you launch, because that baseline is what makes drift visible. A growing gap usually means one of the five failures, so chart the gap every month.
“Treat the CRM as the count of people and GA4 as the map of where traffic came from,” wrote Joshua Jackai, Vancouver web designer.
9. What should a weekly lead source report show?
A weekly lead source report shows one page, grouped by channel, with three numbers each: leads, qualified leads and closed deals. Leads come from form submissions that reached the CRM. Qualified and closed come from the CRM alone. Attribution research found last-click credit can differ sharply from path-based models (Li and Kannan, 2014).

- Rows: channels, including AI assistants
- Columns: leads, qualified, closed
- Touches: first and last side by side
- Source of truth: the CRM for qualified and closed
- Time: 15 minutes a week once built
A report that needs scrolling does not get read in a meeting. GA4 never sees what sales does after the form. Google offers events for those stages, such as qualify_lead (Google for Developers), but most firms I work with report them straight from HubSpot.
Then set attribution with open eyes. Under Data display, Events, Attribution settings, GA4 offers data-driven and last-click models. Two studies, Li and Kannan, 2014 and Anderl et al., 2016, modelled the whole path instead of the final click. That is why this report keeps first and last touch side by side.
“These channels’ relative contributions are significantly different from those found by other currently used metrics,” wrote Hongshuang Li and P.K. Kannan (Journal of Marketing Research, 2014).
So do not let one model settle a budget argument. When first and last touch point at different channels, that difference is the finding. It usually means one channel starts conversations and another closes them, and cutting either one costs you leads a quarter later. Show both columns to whoever owns the budget, and let the pattern over a full quarter decide, not a single week.
Which GA4 lead tracking steps can you do yourself? Filter them
Seven of the nine steps you can run yourself in 10 minutes to an hour each. Two depend on other people: step 5 needs whoever edits your forms, at about half a day per form type, and step 9 needs the CRM owner. Filter the matrix by who does the work to see the time and evidence for each.
| Step | Who does it | Time | Evidence |
|---|---|---|---|
| Decide what to record | Marketing | 20 minutes | Lemon and Verhoef, 2016 |
| Property and data stream | Marketing | About an hour | Google Analytics Help |
| One lead event | Marketing | 30 minutes | Google for Developers |
| UTM naming convention | Marketing | An hour | Google Analytics Help |
| Source into the CRM | Form editor | Half a day per form type | Y Scouts case study |
| AI referrals | Marketing | 15 minutes | Aggarwal et al., 2024 |
| Test end to end | Marketing | 10 minutes per form | Google Analytics Help |
| Watch for drift | Form editor | 20 minutes a month | Utz et al., 2019; Seckler et al., 2014 |
| Weekly report | CRM owner | 15 minutes a week | Li and Kannan, 2014 |
Who does each step and the time are Joshua Jackai’s estimates from client setups, not published benchmarks.
How many of your leads are missing a source? Check in two minutes
This calculator turns three numbers you can read today into one answer: the share of last month’s leads with no usable source, and the gap between GA4 and the CRM. Enter your own counts. Nothing is sent anywhere, and nothing is saved when you leave the page. The defaults are placeholders, not benchmarks.
The defaults are placeholders, not benchmarks. Enter your own counts. The suggested next step is Joshua Jackai’s judgment; prices are from the Jackai Agency price table, November 2026.
Should you set up GA4 lead tracking yourself or hand it off?
Set it up yourself if you have a few forms and own the CRM; the GA4 steps take an afternoon. Hand it off when the site has many forms, someone else runs the CRM, or leadership wants the report next month. Lead Source Reporting is CA$2,600 and two weeks on my November 2026 price table.
The slow parts sit outside GA4: agreeing a naming convention for UTM parameters, adding hidden fields to every form, mapping them to CRM properties, and rechecking forms after plugin updates. Either way, here is what I would not do. I would not count form_submit as the lead event. I would not rely on a hand-picked dropdown as the only record of where a lead came from. I would not add a second analytics tool before the first one is wired end to end. And I would not redesign the site to fix a tracking problem. Tracking runs on the site you already have.
| What you need | Option | Price (CAD) | Time |
|---|---|---|---|
| A few forms, you own the CRM | Do it yourself with the 9 steps | CA$0 | About an afternoon, plus a day per form type |
| You do not know where leads drop off | Website Audit | CA$995 | 1 week |
| Leads arrive with no source | Lead Source Reporting | CA$2,600 | 2 weeks |
Lead Source Reporting is GA4 lead tracking done for you: fixed quote, no design work. You get forms on the pages that already get traffic, plus GA4 and UTM parameters wired into your CRM. First and last touch sit side by side, AI referrals are split out of Direct, and a weekly one-page report names the channel behind each lead.
Questions about GA4 lead tracking, answered
These are the questions owners ask me once the nine steps are in place. They cover why the numbers disagree, what counts as a lead, whether GA4 alone is enough, and the Canadian privacy question. Each answer is short enough to act on today without reopening any of the steps.
Can GA4 track leads on its own?
GA4 can count leads, but it cannot tell you who they are, because Google bars personal information in Analytics. It records the lead event and the channel. Your CRM holds the named contact. Join the two with hidden form fields that pass the same source values into both places.
Why does GA4 show more leads than my CRM?
GA4 usually shows more leads than your CRM because it counts every form, including newsletter signups, as a lead. Count only generate_lead, fired by forms that create a CRM contact. Duplicate Google tags also double the count. A small, steady gap is normal; a growing gap signals a fault.
What counts as a lead in GA4?
A lead in GA4 is whatever event you mark as a key event. I mark only generate_lead, fired when a contact, quote or booking form succeeds. Newsletter signups and downloads can be separate events. Keeping one lead event means GA4 and the CRM count roughly the same people.
Is GA4 lead tracking allowed in Canada?
Yes. GA4 lead tracking is allowed in Canada when you comply with PIPEDA or a provincial law such as BC’s: explain the tracking in plain words and respect a visitor’s choice. Never put names or email addresses in URLs, which Google’s rules forbid. This is not legal advice.
Where to start with GA4 lead tracking this week
Start with step 1 on paper, then step 3, because one clean lead event fixes the most common reason GA4 and the CRM disagree. Then run the phone test in step 7 on your main form. Those three take an afternoon and tell you whether the rest of the chain holds.
I record your current lead count and how many of those leads carry a source at the start of every engagement, so every week after has something to compare against. If you would rather talk it through first, book a 30-minute call. We will measure where your leads come from today, and you leave with that number whether or not you hire me. If AI answer engines are part of the worry, the free AI Visibility Report checks how they describe your firm, within 24 hours.
Sources
Peer-reviewed research
- Aggarwal, P., Murahari, V., Rajpurohit, T., Kalyan, A., Narasimhan, K., and Deshpande, A. (2024). GEO: Generative Engine Optimization. Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining.
- Anderl, E., Becker, I., von Wangenheim, F., and Schumann, J. H. (2016). Mapping the customer journey: Lessons learned from graph-based online attribution modeling. International Journal of Research in Marketing.
- Lemon, K. N., and Verhoef, P. C. (2016). Understanding customer experience throughout the customer journey. Journal of Marketing.
- Li, H., and Kannan, P. K. (2014). Attributing conversions in a multichannel online marketing environment: An empirical model and a field experiment. Journal of Marketing Research.
- Seckler, M., Heinz, S., Bargas-Avila, J. A., Opwis, K., and Tuch, A. N. (2014). Designing usable web forms: Empirical evaluation of web form improvement guidelines. Proceedings of the SIGCHI Conference on Human Factors in Computing Systems.
- Utz, C., Degeling, M., Fahl, S., Schaub, F., and Holz, T. (2019). (Un)informed consent: Studying GDPR consent notices in the field. Proceedings of the 2019 ACM SIGSAC Conference on Computer and Communications Security.
Industry, vendor, government and first-party (not peer reviewed)
- Google Analytics Help and Google for Developers, checked November 2026: About key events, Mark events as key events, Recommended events, Campaign URLs, Default channel group, Custom channel groups, Internal traffic, Unwanted referrals, Data retention, DebugView, Custom dimensions, Personal information.
- Office of the Privacy Commissioner of Canada. The Personal Information Protection and Electronic Documents Act (PIPEDA).
- Jackai Agency price table, November 2026 (first-party prices), and the Y Scouts case study (client HubSpot data).


